What Is an AI Harness? A Practical Guide with a Laravel CMS Example

An AI harness is a structured set of files, rules, skills, workflows, and project context that surrounds an AI coding agent. It provides a consistent way of working across tasks and sessions by defining what must always be true (rules), how specific engineering tasks should be handled (skills), the order in which work should happen (workflows), and facts about the current project (context).

This approach ensures that developers do not have to repeat the same instructions in every prompt, as they are stored within the harness environment.

The key difference between a prompt and a harness lies in their purpose. A prompt is designed for task-specific intent, while a harness is meant for persistent engineering knowledge. By using a harness, developers can store engineering knowledge and project context once, rather than repeating it in each prompt. This leads to a more reliable operating environment as the harness provides discipline to the agent's actions.

An AI harness typically includes several components that serve different responsibilities:

1. AGENTS: Defines roles and authority for the AI coding agent.

2. skills: Reusable engineering knowledge on how to handle specific concerns.

3. rules: Constraints that must always be followed, such as approval gates and the definition of done.

4. workflows: The sequence in which work should occur for a particular type of task.

5. context: Project-specific facts that help guide the agent's actions.

6. adapters: Tool-specific integrations that connect different coding tools to the harness.

For example, in a Laravel CMS project, an AI harness might include files such as AGENTS.md, skills/ (testing strategy, API design, database design, code review), rules/ (core rules, approval gates, definition of done), workflows/ (feature development, bug fixes, release), context/ (project, architecture, domain, conventions), and adapters/ (app, database, tests). This structure helps maintain consistency across sessions and project knowledge, while allowing for adaptable and reusable engineering practices.